Chapter 14
Fronts and Frontogenesis
Satpix 1
Satpix 2
WV Imagery 1200 06/04/2000
Problems with simple frontal models
Chapter 13 examines some simple air mass models of fronts and shows these to have certain deficiencies in relation to observed fronts.
Sawyer (1956) - "although the Norwegian system of frontal analysis has been generally accepted by weather forecasters since the 1920's, no satisfactory explanation has been given for the ‘up-gliding’ motion of the warm air to which is attributed the characteristic frontal cloud and rain. "
"Simple dynamical theory shows that a sloping discontinuity
between two air masses with different densities and velocities
can exist without vertical movement of either air mass...".
Sawyer =>
"A front should be considered not so much as a stable area of strong temperature contrast between two air masses, but as an area into which active confluence of air currents of different temperature is taking place".
y y
cold
warm
1
1
2
2 Several processes including friction, turbulence and vertical motion (ascent in warm air leads to cooling, subsidence in cold air leads to warming) might be expected to destroy the sharp temperature contrast of a front within a day or two of formation.
Clearly defined fronts are likely to be found only where
active frontogenesis is in progress; i.e., in an area where the horizontal air movements are such as to intensify the
horizontal temperature gradients.
These ideas are supported by observations.
Two basic horizontal flow configurations which can lead to frontogenesis:
The intensification of a horizontal temperature gradient by (a) horizontal shear, and (b) a pure horizontal deformation field.
y
x y
x v(x)
isotherms
The kinematics of frontogenesis
Q
O x P
x
u(x +x, t)
u(x, t)
i 1 1
i j 2 i, j j,i 2 i, j j,i j
j call e call
call u ij ij
i, j
u u x [ (u u ) (u u )] x
x
summation over the suffix j is implied
In tensor notation
Relative motion near a point in a fluid
It can be shown that e
ijand
ijare second order tensors e
ijis symmetric (e
ji= e
ij)
ijantisymmetric (
ji=
ij).
ijhas only three non zero components and it can be shown that these form the components of the vorticity vector.
i 1 1
i j 2 i, j j,i 2 i, j j,i j
j call e call
call u ij ij
i, j
u u x [ (u u ) (u u )] x
x
Consider the case of two-dimensional motion
u
1 e
11 x
1 e
12 x
2
12x
2 u
2 e
21 x
1 e
22 x
2
21x
1Note: = =0
Write (x, y) = (x
l, x
2) and (u, v) = (u
l, u
2) and take the origin of coordinates at the point P => (x
l, x
2) = (x, y).
21 u
21 u
12 v
x u
y
12
is the vertical component of vorticity
u
v
u v u
v u v
x y
x x y
x y y
L N
MO Q P L N M
M O
Q P P L
N MO Q P
12 1
2
12 1
2
( )
( )
In preference to the four derivatives v
x, u
y, v
x, v
y, define the equivalent four combinations of these derivatives:
D = u
x+ v
y, called the divergence
E = u
x v
ycalled the stretching deformation F v
x u
ycalled the shearing deformation
v
x u
ythe vorticity
E is called the stretching deformation because the velocity components are differentiated in the direction of the
component.
F is called the shearing deformation because each velocity
component is differentiated at right angles to its direction.
Obviously, we can solve for u
x, v
y, v
x, v
yas functions of D, E, F, .
u
v
u v u
v u v
x y
x x y
x y y
L N
MO Q P L N M
M O
Q P P L
N MO Q P
12 1
2
12 1
2
( )
( )
may be written in matrix form as
u
v
D D
E F F E
x y
F H
GIKJ F
H G IKJ F
H G IKJ F
H G IKJ L N
M O
Q PF H GIKJ
12
0 0
0
0
or in component form as
u u
0
12Dx
12Ex
12Fy
12 y 0 ( x
2)
v v
0
1Dy
1Ey
1Fx
1 x 0 ( x
2)
Then
u = uu
o, v = v v
o, and (u
o, v
o) is the translation velocity at the point P itself (now the origin).
u u
0
12Dx
12Ex
12Fy
12 y 0 ( x
2) v v
0
12Dy
12Ey
12Fx
12 x 0 ( x
2)
Choose the frame of reference so that u
o= v
o= 0
u = u, v = v.
The relative motion near the point P can be
decomposed into four basic components as follows:
(I) Pure divergence (only D nonzero) (II) Pure rotation (only nonzero)
(III) Pure stretching deformation (only E nonzero)
(IV) Pure shearing deformation (only F nonzero).
Pure divergence (I) Pure divergence (only D nonzero)
u
12Dx v ,
12Dy
The motion is purely radial and is from or to the point
u 12 Dr(cos , sin ) 12 Dr
r is the position vector from P.
P P
D > 0 D < 0
Pure rotation (II) Pure rotation (only nonzero).
u
12 y v ,
12 x
u
12 r ( sin , cos )
12 r
The motion corresponds with solid body rotation with angular velocity .
12the unit normal vector to r
r
u
P
y
x
axis of dilatation for E > 0
axis of contraction streamlines
for E > 0
On a streamline, dy/dx = v/u = y/x , or xdy + ydx = d(xy) = 0.
u
12Ex , v
12Ey
The streamlines are rectangular hyperbolae xy = constant.
(III) Pure stretching deformation (only E nonzero)
u
12Fy , v
12Fx
The streamlines are given now by dy/dx = x/y
y
2 x
2= constant.
The streamlines are again rectangular hyperbolae, but with their axes of dilatation and contraction at 45 degrees to the coordinate axes.
y
x 45
oThe flow directions are for F > 0.
(IV) Pure shearing deformation (only F nonzero)
u v
E F F E
x y
F H
GIKJ L
N M O Q PF H GIKJ
12
By rotating the axes (x, y) to (x', y') we can chooseso that the two deformation fields together reduce to a single deformation field with the axis of dilatation at angle to the x axis.
y'
y
x
x'
(V) Total deformation (only E and F nonzero)
Let the components of any vector (a, b) in the (x, y) coordinates be (a', b') in the (x', y') coordinates:
a b
a b
F H
GIKJ F
H G I
K J F H GIKJ
cos sin
sin cos
a b
a b
F
H GIKJ F
H G I
K JF H GIKJ
cos sin sin cos
and
12
u E F x
v F E y
E E F
F F E
cos sin
cos sin
2 2
2 2
where
u v
E F
F E
x y
F H
GIKJ L
N M O Q PF H GIKJ
12
E and F, and also the total deformation matrices are not invariant under rotation of axes, unlike, for example,
the matrices representing divergence and vorticity
12
u E F x
v F E y
E E F
F F E
cos sin
cos sin
2 2
2 2
wher e
We can rotate the coordinate axes in such a way that F' = 0;
then E' is the sole deformation in this set of axes.
tan 2 = F/E E ( E
2 F
2 1 2)
/E'
2+ F'
2= E
2+ F
2is invariant under rotation of axes.
and
u v
E F
F E
x y
F H
GIKJ L
N M O Q PF H GIKJ
12
y
x
axis of dilatation
axis of contraction The stretching and shearing deformation fields may be
combined to give a total deformation field with strength E' and
with the axis of dilatation inclined at an angle to the x- axis.
In summary, the general two-dimensional motion in the neighbourhood of a point can be broken up into a field of divergence, a field of solid body rotation, and a single
field of total deformation, characterized by its
magnitude E' (> 0) and the orientation of the axis of dilatation, .
We consider now how these flow field components act to change horizontal temperature gradients.
General two-dimensional motion near a point
One measure of the frontogenetic or frontolytic tendency in a flow is the frontogenesis function:
D
h/ Dt
D Dt / / t u / x v / y w / z
Start with the thermodynamic equation
D
Dt q
diabatic heat sources and sinks
Differentiating with respect to x and y in turn
The frontogenesis function
D
Dt x
u x x
v x y
w x z
q x
F
H GIKJ
and
DDt y
u y x
v y y
w y z
q y
F
H GIKJ
D
Dt x y
D
Dt x
D
Dt y
|h |
F
, ,H G IKJ
L F H GIKJ FH GIKJ
N M O
Q P
2 2
Now
u D E v F
v D E u F
x x
y y
U V
| W
|
12 1
2
12 1
2
( ), ( ),
( ), ( ),
Use
D
Dt q q w w
D E F E
h x x y y x x y y z
h x x y y
2
2 2 2
2 2 2
2
( )
[ ]
Note that does not appear on the right-
hand-side!
There are four separate effects contributing to frontogenesis (or frontolysis):
h 1 2 3 4
D T T T T
Dt where
T
1 (
x xq
y yq ) /
h n
hq
T
2 ( w
x x w
y y )
z/
h
zn
hw T
3
12D
hT
4
12[ E
x2 2 F
x y E
2y] /
h
unit vector in the
direction of
h
T
1: represents the rate of frontogenesis due to a gradient of diabatic heating in the direction of the existing
temperature gradient
.
T
1 (
x xq
y yq ) /
h n
hq
Heat
Cool q
h
Interpretation
T
2: represents the conversion of vertical temperature gradient to horizontal gradient by a component of differential vertical motion in the direction of the existing temperature gradient
T
2 ( w
x x w
y y )
z/
h
zn
hw
h
hw
T
3: represents the rate of increase of horizontal temperature gradient due to horizontal
convergence (i.e., negative divergence) in the presence of an existing gradient
T
3
12D
h
h
T
4: represents the frontogenetic effect of a (total) horizontal deformation field.
Further insight into this term may be obtained by a rotation of axes to those of the deformation field.
Let denote and relate to T
4
12[ E
x2 2 F
x y E
2y] /
h
x / x
h
h
Solve for E and F in terms of E' and (remember is such that F' = 0)
1 2 2
4 12 h x y x y
2 2
x y x y
T [E cos 2 {( ) cos 2 2 sin 2 }
E sin 2 {( )sin 2 2 cos 2 }] .
y
x
axis of dilatation
axis of contraction
h y`
x`
Schematic frontogenetic effect of a horizontal deformation
field on a horizontal temperature field.
Set
h
h (cos ,sin ) a few lines of algebra
4 12 h
12 h
T E cos 2
E cos 2
The frontogenetic effect of deformation is a maximum when the isentropes are parallel with the dilatation axis ( = 0),
reducing to zero as the angle between the isentropes and the dilatation axis increases to 45 deg.
When the angle is between 45 and 90 deg., deformation has a frontolytic effect, i.e., T < 0.
angle between the axis of dilatation and the potential-temperature isotherms
(isentropes)
x
h
x`
A number of observational studies have tried to determine the relative importance of the contributions T
nto the
frontogenesis function.
Unfortunately, observational estimates of T
2are "noisy", since estimates for w tend to be noisy, let alone for
hw.
T
4is also extremely difficult to estimate from observational data currently available.
A case study by Ogura and Portis (1982, see their Fig. 25) shows that T
2, T
3and T
4are all important in the immediate vicinity of the front, whereas this and other investigations suggest that horizontal deformation (including horizontal
Observational studies
This importance is illustrated in Fig. 14.7, which is taken from a case study by Ogura and Portis (1982), and in Figs.
4.2 and 4.12, which show a typical summertime synoptic
situation in the Australian region.
The direction of the dilatation axis and the resultant deformation on the 800 mb surface at 0200 GMT, 26 April 1979 with the contours of the 800 mb
surface front
From a case study by Ogura and Portis (1982)
H H L
H
A mean sea level isobaric chart over Australia
H H
H
A 1000-500 mb thickness chart over Australia
In a study of many fronts over the British Isles, Sawyer (1956) found that ‘active’ fronts are associated with a deformation field which leads to an intensification of the horizontal temperature gradient.
He found also that the effect is most clearly defined at the 700 mb level at which the rate of contraction of fluid
elements in the direction of the temperature gradient
usually has a well-defined maximum near the front.
Flow deformation acting on
a passive tracer to produce
locally large tracer gradients
from Welander 1955
The foregoing theory is concerned solely with the
kinematics of frontogenesis and shows how particular flow patterns can lead to the intensification of horizontal
temperature gradients.
We consider now the dynamical consequences of increased horizontal temperature gradients
We know that if the flow is quasi-geostrophic, these increased gradients must be associated with increased vertical shear through the thermal wind equation.
We show now by scale analysis that the quasi-geostrophic approximation is not wholly valid when frontal gradients become large, but the equations can still be simplified.
Dynamics of frontogenesis
The following theory is based on the review article by Hoskins (1982).
It is observed, inter alia, that atmospheric fronts are marked by large cross-front gradients of velocity and temperature.
Assume that the curvature of the front is locally unimportant and choose axes with x in the cross-front direction, y in the along-front direction and z upwards:
z
y x
front
L
l
cold warm
V
U
u
hv
u y
x
Frontal scales and coordinates
Observations show that typically, U ~ 2 ms
-1V ~ 20 ms
-1L 1000 km
~ 200 km
=> V >> U and L >> .
The Rossby number for the front, defined as
The relative vorticity (~V/ ) is comparable with f and the motion is not quasi-geostrophic.
Ro V f / ~ 20 ( 10
4 2 10
5)
is typically of order unity.
Du
Dt fv U fV
U V
V / ~
2/
2f
1
F
H GIKJ
and Dv
Dt fu UV fU
V / ~ / f
~
1
A more detailed scale analysis is presented by Hoskins and Bretherton (1972, p15), starting with the equations in
orthogonal curvilinear coordinates orientated along and normal to the surface front.
The motion is quasi -geostrophic across the front, but not along it.
The ratio of inertial to Coriolis accelerations in the x and y
directions =>
The scale analysis, the result of Exercise (14.3), and making the Boussinesq approximation, the equations of motion for a front are
0
zP
D
Dt N w
02 0
fv
xP
Dv
Dt fu
yP
xu
yv
zw 0
P p /
*N
0= the Brunt-Väisälä frequency of the basic state
N
02 ( / g
0)( d
0/ dz )
buoyancy force per unit mass
I assume that f and N are constants.
While the scale analysis shows that frontal motions are not quasi-geostrophic overall, much insight into frontal
dynamics may be acquired from a study of frontogenesis within quasi-geostrophic theory.
Such a study provides also a framework in which later modifications, relaxing the quasi-geostrophic assumption, may be better appreciated.
Quasi-geostrophic frontogenesis
D
Dt t u
x v
y
g
g
g
where v
g= v is computed from fv =
xP as it stands and u
g ( / ) 1 f
yP
Set u = u
g+ u
aDv
Dt fu
a 0
x au
zw 0 and
The quasi-geostrophic approximation involves replacing D/Dt by
0
zP
D
Dt N w
02 0
fv
xP
Dv
Dt fu
yP
xu
yv
zw 0
fv
xP
D v
Dt
g fu
a 0
0
zP
D
Dt
g N w
02 0
x au
zw 0
x gu
y gv 0
fv
xP
0
zP fv
z=
xLet us consider the maintenance of cross-front thermal wind balance expressed by fv
z=
x.
D
Dt
g( fv
z) Q
1 f u
2 azD
Dt
g
xQ
1 N w
02 xNote that ugx + vy = 0
These equations describe how the geostrophic velocity field
acting through Q
lattempts to destroy thermal wind balance by changing fv
zand
xby equal and opposite amounts and how ageostrophic motions (u
a, w) come to the rescue!
Q u v v
gx x x y
x y
1
( , )
( , )
N w
02 x f u
2 az 2 Q
1Also from u
ax+ w
z= 0, there exists a streamfunctionfor the cross-frontal circulation satisfying
( , ) ( u w
a
z,
x)
N
02
xx f
2
zz 2 Q
1This is a Poisson-type elliptic partial differential equation for the cross-frontal circulation, a circulation which is forced by Q
l.
Q
1 u
gx
x v
x
yMembrane analogy for solving a Poisson Equation
2 2
2 2
F(x, y)
x x
F > 0
F < 0
This is an Elliptic PDE Here z = 0 on the domain
boundary
This is called a
Dirichlet condition
2 2
2 2
F(x, y)
x x
F > 0
F < 0
Here = 0 on parts of the domain boundary and /n = 0 on other
parts of the boundary n 0
prescribed on a boundary is called a
Neumann condition.
Slippery glass walls
y
x
u
g= x
v = y Q u v
v
gx x x y
x y
1
( , )
( , ) =
xFrontogenesis in a deformation field
. .
cold warm
z
x
B A
C D
adiabatic warming adiabatic cooling
(northern hemisphere case)
x = 0
Frontogenesis in a field of geostrophic confluence
y
If w = 0, Q
lis simply the rate at which the buoyancy (or
temperature) gradient increases in the cross-front direction following a fluid parcel, due to advective rearrangement of the buoyancy field by the horizontal motion.
Q u v v
gx x x y
x y
1
( , )
( , )
xincreases due to confluence (u
x< 0) acting on this
component of buoyancy gradient and due to along-front horizontal shear v
xacting on any along-front buoyancy gradient
y.
D
x/Dt is an alternative measure of frontogenesis to the Boussinesq form of the frontogenesis function D|
h|/Dt analogous to the left hand side of this, i.e., T
1+ T
2+ T
3+ T
4.
D
Dt
g
xQ
1 N w
02 x1 gx x x y
Q u v
The quasi-geostrophic theory of frontogenesis in a field of pure geostrophic deformation was developred by Stone (1966), Williams and Plotkin (1968), and Williams (1968).
The solutions obtained demonstrate the formation of large horizontal gradients near boundaries, but away from
boundaries, the induced ageostrophic circulation prevents the contraction of the horizontal length scale of the
temperature field below the Rossby radius of deformation, L
R= NoH/f; where H is the depth of the fluid.
Because the ageostrophic circulation does not contribute to advection in quasi-geostrophic theory, the largest
horizontal temperature gradient at each height remains
coincident with the line of horizontal convergence (x = 0).
Limitations of quasi-geostrophic theory
Many unrealistic features of the quasi-geostrophic theory result from the omission of certain feedback mechanisms.
The qualitative effect of some of these feedbacks can be
deduced from the quasi-geostrophic results.
. .
cold warm
z
x
B A
C D
x = 0
The ageostrophic velocity u
ais clearly convergent (u
ax< 0) in the vicinity of A on the warm side of the maximum T
x(
x).
If included in the advection ofit would lead to a larger
gradient .
. .
cold warm
z
x
B A
C D
x = 0
At A, the generation of cyclonic relative vorticityis
underestimated because of the exclusion of the stretching term
wzin the vertical vorticity equation,
D
Dt f w
z
( )
Similar arguments apply to the neighbourhood of C on the cold side of the maximum temperature gradient at upper levels.
In the vicinity of B and D, the ageostrophic divergence would imply weaker gradients inand the neglect of
wz. .
cold warm
z
x
B A
C D
x = 0
D
Dt ( f ) w
z In summary, QG-theory points to the formation of sharp
surface fronts with cyclonic vorticity on the warm side of the temperature contrast, and with the maximum horizontal
temperature gradient sloping in the vertical from A to C, even though these effects are excluded in the QG-solutions.
The theory highlights the role of horizontal boundaries in
frontogenesis and shows that the ageostrophic circulation acts to inhibit the formation of large gradients in the free
atmosphere.
Hoskins (1982) pointed out that unless the ageostrophic convergence at A and C increase as the local gradients
increase, the vorticity and the gradients incan only increase exponentially with time.
Quasi-geostrophic theory does not even suggest the formation
of frontal discontinuities in a finite time.
Semi-geostrophic frontogenesis
The so-called semi-geostrophic theory of frontogenesis is obtained from the unapproximated forms of the frontal equations:
in other words, we do not approximate D/Dt by D
g/Dt and
therefore advection by the total wind is included.
0
zP
D
Dt N w
02 0
fv
xP
Dv
Dt fu
yP
xu
yv
zw 0
fv
xP
D v
Dt
g fu
a 0
0
zP
D
Dt
g N w
02 0
x au
zw 0
x gu
y gv 0
fv
xP
0
zP fv
z=
xy
f of Dv fu P
z Dt
As before, cross-front thermal-wind balance fv
z=
xD
Dt ( fv
z) Q
1 F u
2 az S w
2 zF
2 f f v (
x) S
2 fv
z
xD
Dt
x Q
1 u
ax x w N
x 22
2
Now
2 o
of D N w 0
x Dt
also
D
Dt ( fv
z) Q
1 F u
2 az S w
2 z2
x 1 ax x x
D Q u w N
Dt
is the total Brunt-Väisälä frequency, rather than that based on the basic state potential temperature distribution.
N
2 N
02
zTo maintain thermal-wind balance ( fv
z=
x)
N
2
xx 2 S
2
xz F
2
zz 2 Q
12 2 2
1 ax x x 1 az z
Q u w N Q F u S w
N
2
xx 2 S
2
xz F
2
zz 2 Q
1This is the equation for the vertical circulation in the semi- geostrophic case.
It is elliptic provided that the so-called Ertel potential vorticity,
This condition which ensures that the flow is stable to symmetric baroclinic disturbances as discussed in a later course (Advanced Lectures on Dynamical Meteorology).
1 2 2 4
q f (F N
S ) 0
N
02
xx f
2
zz 2 Q
1Compare with the QG-circulation equation
z
X
z
x X
1X
2x
1x
2(a) The circulation in the (X, Z) plane in a region of active frontogenesis (Ql > 0). (b) The corresponding circulation in (x,z)-space. The dashed lines are lines of constant X which are close together near the surface, where there is large cyclonic vorticity.
X = x + v
g(x,z)/f
(a) (b)
y
x
u
g= x v = y
i(x) 2 tan
1x L
Frontogenesis in a deformation field
= 12
oC
H H
H